Objective To clarify the effect of postoperative (adjuvant) external-beam pelvic radiotherapy (EBRT) for different grades of early endometrial cancer.
Search strategy Meta-analysis of data from randomised trials stratified by histological risk factors supported by cohort studies.
Selection criteria Cochrane methodology.
Data Seven randomised trials were identified. Five were eligible for meta-analysis. Homogeneity was confirmed (I2 < 25%).
Main outcome measures Survival, site of recurrence and added complications.
Main results EBRT after hysterectomy for low-risk disease increases the odds of death (OR for overall survival 0.71; 95% CI 0.52–0.96). EBRT does not appear to alter survival for intermediate-risk cancers (stage ICG1/2 and IBG3) (OR 0.97; 95% CI 0.69–1.35). In contrast, EBRT offers a significant disease-free survival advantage for high-risk cancer (OR 1.76; 95% CI 1.07–2.89). The survival advantage benefits one in ten women. The definition of high risk is variable across studies but focuses on ICG3 (deeply invasive, poorly differentiated) tumours. Pelvic EBRT reduces the risk of pelvic recurrent disease in all types of invasive endometrial cancer (OR 0.27; 95% CI 0.16–0.44), but local recurrence may respond to salvage treatment. The risk of distant metastasis appears to be increased significantly by prophylactic EBRT (OR 1.58; 95% CI 1.07–2.35), but this might be because pelvic relapse in untreated women alters reporting of metastatic disease.
Authors’ conclusions Adjuvant EBRT should not be used for low- (IA, IBG1) or intermediate-risk (IBG2) cancer, but it is associated with a 10% survival advantage for high-risk (stage ICG3) endometrial cancer. This challenges the role of a staging lymphadenectomy.
Endometrial cancer is an increasing public health problem in developed countries. More than 1 in 20 cancers affecting women in Europe involve the endometrium.1 It is the fifth commonest cancer in women,2 and there are 38 000 cases annually in the USA.3 Recent UK surveys have shown that the 60- to 79-year female age group incidence of endometrial cancer has climbed from 48 cases per 100 000 in 1993 to 63 in 2001, a rise of 30% in less than a decade.4 Most endometrial cancers are diagnosed with stage I disease.3 Stage I endometrial cancer defines a cancer that has not spread outside the uterus. Stage IA is cancer confined to the inner layer of cells of the uterus (endometrium). Stage IB is cancer that invades less than one half of the muscle wall of the uterus. Stage IC is cancer that invades more than one half of the muscle wall of the uterus. Primary curative treatment of early-stage endometrial cancer is hysterectomy and bilateral salpingo-oophorectomy. The site of first relapse in surgically staged patients is usually the upper vagina, or ‘vaginal vault’,5,6 and this can be reduced by adjuvant postoperative radiotherapy.5
Several postoperative treatment options are currently promoted: a surveillance policy, adjuvant pelvic external-beam radiotherapy (EBRT), and adjuvant vaginal vault brachytherapy (short-range radiotherapy). A surveillance policy is designed to spare the cost and toxicity of adjuvant treatment, keeping radiotherapy in reserve to salvage any pelvic relapses of the cancer that might occur. Adjuvant pelvic EBRT is designed to irradiate sites of potential micrometastatic local cancer spread: the upper vagina (the cervix cuff excision margin, or vaginal vault), parametrial ligaments, and primary draining lymph nodes. Treatments typically involve 25–28 EBRT treatments delivered over 5–6 weeks of outpatient hospital treatment. Vaginal vault brachytherapy is designed to treat only the upper vagina. Treatment typically involves one to five internal vaginal radiotherapy treatments, irradiating a much smaller volume of the patient than with EBRT. Despite the prevalence of this disease, there is no clear consensus guideline on the relative merits of these three approaches for patients with early-stage endometrial cancer after a hysterectomy.7 Attempts to decide a cost–benefit relationship for adjuvant treatment have been difficult because there has been no systematic review of the evidence for the effect of these adjuvant treatment strategies on overall survival.8 New data from Surveillance, Epidemiology, and End Results (SEER)9 and five randomised trials have recently been presented at scientific meetings. New randomised trial data come from the American GOG 99 trial,10,11 Japanese GOG 2033 trail,12 Argentinean high-risk trial,13 and UK ASTEC trial,14 and we now have the 10-year survival data15 from the Dutch PORTEC trail.16 These data can be added to the Norwegian Radium Hospital trial17 and the American trial of preoperative radiotherapy18 to systematically study the effect of adding pelvic EBRT to hysterectomy.
Simple amalgamation of data would fail to recognise that endometrial cancer is not one disease. It is intuitively logical that a potentially dangerous treatment will not help women if the risk of recurrence is lower than the potential for harm. The recurrence risk from an IBG1 (minimally invasive, well differentiated) tumour is about 2%, and radiotherapy is inappropriate for this low-risk cancer. However, it might be appropriate for intermediate- or high-risk cancer. Therefore, this analysis separates cases of endometrial cancer into different stages and grades of disease.
The full methodology in this meta-analysis is described by the Cochrane collaboration user guide.19 The search strategy copied the protocol in the Cochrane Database of Systematic Reviews.20 The trials are well known in expert circles, and the existence of others was explored using the UK Cochrane trial register, computer databases, hand-search, and cascade searching. All randomised trials studying prophylactic radiotherapy for excised endometrial cancer were included. The primary objective was to compare the survival and disease-free survival from primary endometrial endometrioid adenocarcinoma with and without adjuvant EBRT. Only randomised trials considering International Federation of Obstetrics and Gynecology stage I and II disease treated by primary hysterectomy and bilateral salpingo-oophorectomy were considered for this systematic review and meta-analysis. The review excluded the trials comparing radiotherapy with chemotherapy, cancer beyond the uterus, and serous papillary pathology. The search strategy for identification of studies followed the Cochrane Gynaecological Cancer Collaborative Review Group search strategy.21 This uses the following databases: MEDLINE (1951, 16 March 2005), EMBASE (1974, 16 March 2005), Cochrane Central Register of Controlled Trials (CENTRAL) on the Cochrane Library, Cochrane Database of Systematic Reviews, Gynaecological Cancer Group Specialised Register, PubMed, TRIP, Trials Central, Current Controlled Trials, and Centerwatch Clinical Trials Listing Service. Other data relating to results from the PORTEC study were obtained directly from the authors. The risk of recurrence and survival were extracted from the papers, and the meta-analysis was performed with shareware Rev Man 4.2.22 The ratio of events in the two treatment arms were calculated for meta-analyses of dichotomous outcomes. The generic inverse variance method was used for time to death and time to recurrence. Adverse events related to therapy were abstracted. Statistical heterogeneity was assessed using the I2 and Q statistic. The magnitude of statistical heterogeneity was assessed using the I2 statistic. A fixed-effects model was used after confirming there was no evidence of statistical heterogeneity (prospectively set as I2 < 25%).
Seven completed randomised trials were identified that focused on the role of pelvic EBRT for stage I endometrial cancer treated by hysterectomy and salpingo-oophorectomy. Two were excluded. Data from the radiation arm of the ASTEC trial are not sufficiently mature for inclusion, and the Japanese GOG 2033 study compared radiotherapy with chemotherapy, not adjuvant radiotherapy versus no radiotherapy. The five suitable trials involve different populations with different treatments, but all focus on the survival advantage of EBRT (Table 1). There is no statistical heterogeneity to complicate the interpretation of data (I2 < 25% in each prospectively selected analysis). Women in the American GOG 99 multicentre trial and Argentinean trial had a pelvic lymphadenectomy, and extrauterine metastases were exclusion criteria. Women in the Argentinean trial had high- and intermediate-risk cancers (IBG2-3, IC), whereas the Dutch multicentre PORTEC trial specifically excluded ICG3 cancer. Women in the Norwegian Radium Hospital trial all had radioactive caesium applied to the vaginal epithelium after hysterectomy. A smaller American trial of preoperative radiotherapy randomly allocated women a single implant with Heyman (uterine cavity brachytherapy) capsules possibly with tandem and ovoids, or external megavoltage irradiation before hysterectomy.
Table 1. Characteristics of the included trials
Type of trial
Pelvic radiation regime
Complications attributable to pelvic radiation
Stage 2 and ICG3 were excluded
All but one of the 20 Dutch centres were involved
Women were randomly allocated 46 Gy using 2-Gy fractions 5 days per week or no additional therapy
A total of 339 women started radiotherapy. Follow-up reports confirmed nine cases of grade 3 gastrointestinal late toxicity requiring surgery. Three needed surgery for sigmoid stenosis and six had small-bowel surgery. One died of an exacerbation of Crohn’s disease soon after radiotherapy
All women received a staging lymphadenectomy to exclude node metastasis
Multicentre NCI grant to the GOG
Women were randomly allocated 50.4 Gy given as 28 or more fractions of 0.18 Gy to the whole pelvic beginning within 8 weeks of surgery or no additional treatment
A total of 190 women received radiotherapy, two died of related intestinal damage, 15 women suffered grade 3 or 4 intestinal toxicity compared with 2 in the surgery-only arm
Hysterectomy and bilateral oophorectomy followed 4–8 weeks of radiotherapy. No information is available on tumour grade
Women were randomly allocated intracavity multiple Heyman capsules and tandem and ovoids with caesium 137 or external-beam irradiation giving 4000 rad over 4 weeks
A total of 55 women received intracavity radiation and there were six major complications (two cases of fatal acute hepatic necrosis, operative injury to ureter causing fistulae, wound dehiscence, viral hepatitis and a case of severe radiation cystitis). Fifty women received external-beam radiation, and there were five serious complications (two cases of severe radiation enteritis, small-bowel infarction, bowel perforation, and a case of radiation fistulae)
IBG2-3, IC only, and all women received a staging lymphadenectomy to exclude cases of node metastasis
Women were randomly allocated 5000 cGy to the whole pelvis or no additional treatment
Women with stage I (Mo Ro) endometrial cancer treated by median incision, hysterectomy, and bilateral salpingo-oophorectomy (with out lymphadenectomy) and intravaginal radium delivering 6000 rad to the vagina
All women had 6000 rad from intravaginal radium, and then women were randomly allocated, no additional therapy or 4000 rad in 20 fractions over 4 weeks to the pelvic lymph nodes with central shielding at 2000 rads with a boost to the external iliac vessels of 300–400 rad via an intravaginal radium applicator
A total of 263 women received radiotherapy. Two died of pelvic radiotherapy complications. One died of unspecified complications, the other had ileal resection for obstruction and adhesions complicated by fistulae and death from sepsis. One other woman required bladder resection for radiation necrosis
Women with low- or intermediate-grade cancers were more likely to die if they received prophylactic EBRT (OR for overall survival 0.71; 95% CI 0.52–0.96). This is statistically and clinically significant (Figure 1). The crude data suggest that EBRT was associated with one extra fatality per 30, and we can be at least 95% certain that prophylactic pelvic EBRT is either harmful or ineffective in improving survival in women with low- or intermediate-risk cancers. This conclusion is robust even if data from the preoperative trial are ignored (OR 0.73; 95% CI 0.52–1.01).
EBRT does not appear to affect overall survival in intermediate-risk endometrial cancer, defined as either stage ICG1 (deeply invasive, well differentiated) or stage IBG3 (superficially invasive, poorly differentiated) tumours. Disease-free survival in women with good World Health Organization performance status (2 or less) is similar in women who did not receive adjuvant EBRT (OR 0.85; 95% CI 0.59–1.24) (Figure 2). These data are limited to the overall 10-year disease-free survival data from the PORTEC trial from women older than 60 years of age, and 5-year survival data from the Radium Hospital study (all ages). The addition of all high-risk data from GOG 99 to this analysis does not favour EBRT (OR 0.97; 95% CI 0.69–1.35). It is noteworthy that death specifically from high-intermediate-risk endometrial cancer in the PORTEC trial was lower in women who received EBRT (12% compared with 14%), but deaths from all causes at 10 years were higher after EBRT (actuarial percentage 39.2% compared with 36.5%).
This negative effect on survival from adjuvant radiotherapy does not apply to high-risk cancer. Meta-analysis of high-risk cases from the GOG 99 and Radium Hospital trial suggests that prophylactic EBRT might be associated with a 10% absolute reduction in death from cancer (OR 0.59; 95% CI 0.30–1.17). In other words, ten women have to accept a 3% risk of severe complications and a 20% risk of mild (mostly grade 1) long-term toxicity that affects their quality of life23,24 for one to gain a potential cure from prophylactic postoperative pelvic radiotherapy. The limitation of restricting the analysis to cancer deaths is that a 10% reduction in death from cancer may not be overall survival advantage. Half the women in this group in GOG 99 died due to causes unrelated to their cancer, so the 10% apparent long-term survival advantage will be diluted by deaths from other causes. The disease-free survival is more useful, and data from the Argentinean trial can be included in this analysis. The Argentinean report is limited to crude numbers, and the high- and intermediate-risk cancers are not separable. However, inclusion of these data to the meta-analysis emphasises the disease-free survival advantage from 69 to 80% associated with prophylactic radiotherapy (OR 1.76; 95% CI 1.07–2.89) (Figure 3).
It is also important to note that pelvic EBRT reduces the risk of pelvic recurrent disease in all types of invasive endometrial cancer (Figure 4) (OR 0.27; 95% CI 0.16–0.44). This risk reduction does not translate to improved survival from lower risk cancers. Most women do not develop recurrent disease if prophylactic radiotherapy is withheld, and recurrent disease in the pelvis may respond to later salvage radiotherapy. In contrast, pelvic recurrence after prophylactic postoperative radiotherapy is usually fatal. The risk of distant metastasis does not seem to be reduced by prophylactic radiotherapy (Figure 5) (OR 1.58; 95% CI 1.07–2.35). The final major consideration before prescribing pelvic radiotherapy is serious toxicity. Fatalities from radiotherapy were reported in these trials (table), and there are other management options. Vaginal vault brachytherapy may offer high rates of cancer control with low risks of toxicity.25
These data show that prophylactic postoperative pelvic EBRT for low-risk cancers is harmful and has no long-term survival advantage. It will reduce the risk of local recurrence but not the risk of distant metastasis or overall survival. In contrast, EBRT improves overall survival and recurrence-free survival from high-risk tumours. This conclusion is supported by observations from cohorts who were offered prophylactic EBRT arbitrarily. There was no survival benefit from postoperative EBRT for 724 women with registered stage I endometrial adenocarcinoma treated by hysterectomy between 1975 and 1992 in Eindhoven when radiotherapy was provided according to the different referral patterns between different hospitals. The majority of cases had low-risk disease. In fact, the treated group had an excess death rate over the normal Dutch female population, and this did not decrease during the 10-year follow up.26 The Swedish Council of Technology Assessment in Health Care reported observations from 3446 cases and offered further evidence that adjuvant EBRT has no value in low-risk disease.27 The SEER study is a retrospective analysis of 21 249 women, including data from the SEER program of the US National Cancer Institute from 1 January 1988 to 31 December 2001.28 Adjuvant EBRT does not appear to offer any benefit and may reduce survival in low-risk disease, but it was significantly associated with an improved overall survival from stage ICG3 endometrial adenocarcinoma (hazard ratio 0.72; 95% CI 0.57–0.92).
Interpreting cohort studies is difficult because of bias. In other words, the results from cohort studies might be due to a variety of confounding variables such as a trend to omit radiotherapy when women have a limited life expectancy from other pathology. Randomised trials avoid this. Randomised trial data confirm that prophylactic postoperative EBRT will prevent some cases of recurrent pelvic disease in a small proportion of women. However, it makes more sense to advise a surveillance policy for low-risk disease because of the potential for toxicity from EBRT and a worse overall survival. In the PORTEC study, 75% with pelvic recurrence could be treated with curative intent and 85% achieved complete remission. The 2- and 3-year survival rate after relapse was 79 and 69%, respectively, in the control group compared with 21 and 13% in women treated by prophylactic EBRT. The rate of distant metastatic cancer was similar in both groups, implying that there is a subgroup of rapidly metastasising cancers that spreads to distant sites before the initiation of radiotherapy. Prophylactic radiotherapy will not cure these women, but it will harm them. Prophylactic radiotherapy does reduce local recurrence, but selective salvage therapy may be just as effective and vaginal brachytherapy may be adequate prophylaxis for local recurrent disease. There are multiple retrospective case series to support this policy.25,29–34 This issue is being explored prospectively in the current Dutch PORTEC 2 trial.35
There are obvious advantages to omitting postoperative pelvic EBRT for low- and high-intermediate-risk cancers. The vast majority who are cured by surgery are spared the toxicity, inconvenience, and expense of unnecessary therapy. In contrast to low-risk and high-intermediate-risk cancers, a good case can be made to support prophylactic postoperative EBRT for high-risk cancer. If there is no significant coexisting morbidity, trial data allow us to be more than 99% confident that postoperative EBRT will improve both overall and disease-free survival. This may benefit one woman in ten. This potential 10% survival advantage may make routine staging lymphadenectomy redundant. Radiotherapy in this group will follow hysterectomy irrespective of the pelvic node staging. This survival advantage estimate will be more precise in 5 years, when the mature data from the NCIC Clinical Trials Group36 and MRC (UK) ASTEC37 are available. Long-term outcome data are important as pelvic relapses occur late, and after a relapse, median survival times in excess of 3 years are reported.38
There are obvious limitations to meta-analysis. The data involve heterogeneous cohorts, and subgroup analysis is ad hoc. The randomised trials studied span 25 years of research, and the pathology reporting of endometrial carcinoma operative specimens has changed over this time. Clear cell adenocarcinoma and uterine papillary serous carcinoma are subtypes of poorly differentiated endometrial cancer that may show different patterns of spread and would now be excluded from current trials comparing surveillance policies with adjuvant treatment. Over this period, the aetiology of endometrial cancer has evolved due to the changing use of oestrogen hormone replacement therapy, increasing tamoxifen use and increasing lifespan and obesity. However, all the evidence suggests that women with low-risk tumours should not have EBRT. Data from intermediate-risk cancers are limited by a relatively small sample size but suggest that the disadvantages of adjuvant EBRT may balance the advantages. However, the data from high-risk cancers suggest a survival advantage that cannot be ignored.
This leaves the surgeon wondering if a small risk of lymph node metastasis is worth detecting in intermediate-risk cancers. A diagnostic lymphadenectomy will triage intermediate-risk women with rare micrometastasis to radiotherapy, but the chance of doing good from finding microscopic metastatic deposits may exceed the risk of surgical harm in some cases. In the PORTEC trial, 10 out of 360 women later developed pelvic sidewall recurrence, implying that the risk of undiagnosed lymph node metastasises at the time of hysterectomy was below 10%. The quality of lymph node harvests and pathological appraisal is variable in the UK,39 but for the sake of argument, let us assume that micrometastatic disease is found in half of those who have it. We can take the extreme position and assume that every potentially fatal relapse would have been cured by earlier treatment and assume that as many as 20% with isolated pelvic and vaginal recurrence fail salvage radiotherapy. This may overexaggerate the true value of lymph node staging because only 7 women in 51 died from local recurrent disease in the control arm of the PORTEC trial, and the risk of distant metastasis was similar irrespective of prophylactic therapy (7.0% in the control arm versus 7.9 at 5 years after adjuvant EBRT). These assumptions suggest that the maximal possible survival advantage from removing nonenlarged nodes must be less than 1 in 360. This is because a staging lymphadenectomy will only find half of the cases from an incidence of 1 in 36, and 80% are cured later (20% of half of 10/360). It is difficult to support diagnostic lymph node surgery on 360 women for one to potentially benefit especially when a lymphadenectomy increases permanent radiation lymphoedema risks, and there is a greater treatment-related death rate with lymphadenectomy in the ASTEC trial. A preliminary report from the UK ASTEC trial showed eight treatment-related deaths from 704 women randomly allocated lymphadenectomy compared with 2 from 704 who did not have lymph node surgery.40 The data from ASTEC needs to mature, but current analysis reveals more deaths (103/704) associated with lymphadenectomy compared with 88/704 allocated no lymph node surgery41 (OR 1.20; 95% CI 0.88–1.63), and more cases of detected recurrent disease (144/704 compared with 107/704, OR 1.43; 95% CI 1.09–1.89).
In summary, these data suggest that routine staging lymphadenectomy and preoperative myometrial assessment is unlikely to alter the treatment policy of a typical patient. A potential survival advantage from a stage ICG3 tumour for about one woman in ten makes pelvic EBRT attractive irrespective of node status. Women with low or intermediate cancer may be best treated by hysterectomy only or with adjuvant vault brachytherapy. Pelvic EBRT will reduce the risk of pelvic recurrence from intermediate-risk cancers, but the toxicity is difficult to justify when there is no appreciable survival advantage.
We are grateful to Carien Creutzberg and Astrid Scholten for providing raw data from PORTEC 1. We are grateful to the Department of health (UK) for funding the gynaecological oncology Cochrane collaboration in Bath (UK) and for access to this library and statistical expertise. Each author has participated sufficiently in the work to take public responsibility for appropriate portions of the content. Nick Johnson has full access to all the data and statistics in this paper and takes responsibility for the accuracy of the data analysis and for the integrity of the work as a whole, from inception to published article.